issues about the code

[bowenroom]

Hi! Thanks for releasing the code. I re-implement the code and found that the result is not as good as what the paper has demonstrated. There is a gap between two results especially the car. I follows the step Readme has pointed out. Can you release the data used in the experiment? (data in baidu yun are the not sliced ones, I think they are not the data used in the experiment)

[twsha]

How to visualize and save test set data，When I run the visual command, I get the following error:

[h1063135843]

How to visualize and save test set data，When I run the visual command, I get the following error:

You can try to run pip install -e .. In your screeshot, you use original mmseg, which doesn't support four channel image.

[twsha]

How to visualize and save test set data，When I run the visual command, I get the following error:

You can try to run pip install -e .. In your screeshot, you use original mmseg, which doesn't support four channel image.

thanks ，I would try . I use the command python tools\test.py configs\edft\segformer_mit_fuse-b0_256x256_80k_vai.py mit_fuse_b0.pth --eval mIoU --show ，then an error about the number of channels was reported.

[twsha]

How to visualize and save test set data，When I run the visual command, I get the following error:

You can try to run pip install -e .. In your screeshot, you use original mmseg, which doesn't support four channel image.

I seemed to have run this command at first，but it still didn’t work.

[twsha]

How to visualize and save test set data，When I run the visual command, I get the following error:

You can try to run pip install -e .. In your screeshot, you use original mmseg, which doesn't support four channel image.

and now it raise this error

[twsha]

How to visualize and save test set data，When I run the visual command, I get the following error:

You can try to run pip install -e .. In your screeshot, you use original mmseg, which doesn't support four channel image.

How to enter the mmsegmengtation package in the conda environment


Or run the pip install -e . command directly in the created virtual environment?
do you know how to solve it,thanks.

[dabendanjjbang]

Hello, I directly downloaded your code and dataset and ran it, I found that the error did not come to pretrain/mit_b0.pth, so I went to segformer and downloaded the mit_b0.pth, in the case that I did not change any parameters, I ran python tools\train.py configs\edft\segformer_mit_fuse_b0_256x256_80k_vai.py, on both 1050ti and 3090 are more than 71, is it a problem with the pre-trained model or something else, please guide.

[twsha]

Hello, I directly downloaded your code and dataset and ran it, I found that the error did not come to pretrain/mit_b0.pth, so I went to segformer and downloaded the mit_b0.pth, in the case that I did not change any parameters, I ran python tools\train.py configs\edft\segformer_mit_fuse_b0_256x256_80k_vai.py, on both 1050ti and 3090 are more than 71, is it a problem with the pre-trained model or something else, please guide.

hi,Can you give me your contact information? I want to communicate with you.

[dabendanjjbang]

+v cx3245299704

[twsha]

How to visualize and save test set data，When I run the visual command, I get the following error:

You can try to run pip install -e .. In your screeshot, you use original mmseg, which doesn't support four channel image.

hi , I have clone the EDFT again ,but when I run the command to show ,it's still report this error ,can you tell me how to solve it !thanks!

[h1063135843]

How to visualize and save test set data，When I run the visual command, I get the following error:

You can try to run pip install -e .. In your screeshot, you use original mmseg, which doesn't support four channel image.

hi , I have clone the EDFT again ,but when I run the command to show ,it's still report this error ,can you tell me how to solve it !thanks!

Can you try out the branch "merge to master"? You should make sure mmseg official repo running well.

[twsha]

How to visualize and save test set data，When I run the visual command, I get the following error:

You can try to run pip install -e .. In your screeshot, you use original mmseg, which doesn't support four channel image.

hi , I have clone the EDFT again ,but when I run the command to show ,it's still report this error ,can you tell me how to solve it !thanks!

Can you try out the branch "merge to master"? You should make sure mmseg official repo running well.

Sorry, I would like to ask you again. I used the latest branch to successfully run the segmentation result image. The saveable image is a mask image. Do you know where the configuration needs to be changed? Thank you.